Graphics Group

Computer graphics is notoriously compute-hungry. Yet the ambition of synthesizing all known or imaginable visual effects in real-time remains far from fully realized and will require even more computation. Fortunately, specialized graphics hardware is becoming more powerful. Multi-core computers with many processors are also developing. The amount of computation that can be done in parallel and per memory reference both continue to grow. Efficient graphics algorithms must therefore be computed in parallel, respect the memory hierarchy, and minimize communication between threads and between the CPU and graphics co-processor (GPU). Hardware trends favor algorithms that perform computation only over a small, local working set, stream data from input to output, and amplify data as it moves through memory hierarchy levels successively closer to the computational units. We are exploring new graphics representations and algorithms that take advantage of existing and upcoming hardware features to heighten the quality of real-time computer graphics.

---

People

Primary Contact: Jim Blinn
Blinn, Jim	Photo Not Available Guenter, Brian	Hoppe, Hugues	Loop, Charles	Snyder, John

Projects (Graphics Projects Page)

• Procedural Geometric Representations
  Our research looks at rendering and processing parametric, procedurally-defined representations directly.
  
• The Polynomial Catalog
  We build on graphical notation originally proposed by Sylvester and Clifford over 100 years ago to classify, in terms of a polynomial’s coefficients, the geometric shapes resulting from its solution set.
  
• Direct GPU Rendering of Piecewise Algebraic Curves and Surfaces
  We develop techniques for direct rendering on the GPU of curves and surfaces defined implicitly as solutions to polynomials.
  
• Approximating Subdivision Surfaces for Hardware Tessellation
  To accelerate the rendering of high-order surfaces on the GPU, we exploit a new programmable tessellator unit proposed for future graphics chips and already shipping on the Xbox 360.
  
• Efficient Processing of Sampled Geometric Surfaces and Signals
  We investigate parameterization techniques to represent surface signals and even geometry via “as small as possible” texture maps.
• Random-Access GPU Data Structures
  We introduce methods for packing sparse data into compact tables while retaining efficient random access.
• 3D Surface Acquisition and Reconstruction
  We propose methods and representations to improve the sampling and reconstruction of scanned 3D objects.
  
• Real-Time Texture Synthesis
  We introduce fast methods for synthesizing arbitrarily large textures from small examples on the GPU.
• Real-Time Soft Global Illumination
  We exploit precomputation and low-frequency numerical methods to render soft effects such as shadows from large light sources, subsurface scattering, and participating media in real time.